6 out of 10 people prefer online video platforms to live TV, and 8 out of 10 18-49 year-olds watch YouTube 1. Among the 50 YouTube accounts with the more reach in Spain, only 4 are women, 2 are formed by men and women, and 3 are institutions 2.
The online platforms are structured in a way that impact gender power relation and impose gender roles, creating pressure and expectations. Most women feel unrepresented in online participation spaces as Youtube, and become passive participants 2.
A platform with so many visitors must be gender-inclusive, while we, the users, must hack the search algorithm. We need the presence of women to encourage all genders to create content and be anyone they want.
What it does
Genderhack invites everyone to search for YouTube content in its searcher, so we can start to make more visible all the women in the industry instead of being guided by an algorithm that systematically puts men above women.
How it's built
We have used python for the creation of the gender recognition model. We used the sklern library to archive that. The database is publically accessible and contains male-female labeled voice data.
To serve the search results, we created a flask app which serves an endpoint that takes a search query as body and returns the YouTube videos that have been categorized as created by women. If no videos are available using that query, the backend sends to the model the first 100 YouTube results to classify. Videos classified as women are added to a mySQL database hosted in google cloud. The backend is deployed in a google app engine instance.
We developed a website to allow the users to search for their videos and retrieve the list of videos made by women. As of now, the database is scarce, but the more people use it, the bigger the database will become. The frontend is deployed in another google app engine instance.
Challenges we ran into
In the end we were not able to connect the database to the frontend, so the users are not able to see any of the classified videos :( Also, the backend stopped working and the connection between the endpoint and the gender classifier was lost minutes before the deadline.
Accomplishments that we're proud of
We have successfuly been able to analyze and classify voices and assert the gender of the speaker. We have been able to deploy a flask app to google cloud, which worked as expected for most of the time. We have built a great design which you will love, including the frontend and the video submission.
What we learned
A lot. Python, web developement, supervised learning, using google cloud console... You name it!
What's next for genderhack
We would like to see in action! Really only a couple of details are missing, but the project is almost finished and would be a good place where to find resources made by women.